[time-nuts] Thunderbolt Supply

[Charles P. Steinmetz]

Bob wrote:

>If all you are doing is running a Thunderbolt, you don't need a 
>supply that's more quiet than most batteries.

Most batteries are *very* quiet -- it takes heroic measures to get 
*any* actively regulated supply into that ballpark.  Indeed, one 
might be tempted to run a Tbolt off of three batteries, each one 
charged by a low-noise, high-impedance current source that puts out 
about .05 CV more than the Tbolt draws.  One could even turn the 
charging off for short periods of "minimal noise" operation, if the 
batteries were suitably sized.  However, in either case I would be 
concerned that the drift of one or more of the battery voltages (poor 
absolute regulation) might introduce another source of XO drift -- 
but I have not tried it.

>The idea is to stop spending money when you have reached the "good 
>enough" point.

I quite agree with you on this point (maybe I'd say "when you are 
clearly past the point where other errors dominate the performance 
envelope").  However:

>People get reasonable performance off of straight switcher outputs. 
>Adding simple linear + filtering gets you well into the overkill 
>region on this application.

I'm not sure we know for certain how quiet is "good enough" for a 
Tbolt, or where the "overkill region" is -- particularly when the 
residual noise contains impulse hash from a switching regulator.  I 
presume most time nuts would consider a 5 dB improvement in phase 
noise worthwhile for the relatively low cost/effort of using a good 
linear supply, if that level of improvement can be attained.  The 
only systematic study I've seen is what tvb has on his web site, and 
it does not include data from a switching supply with external 
post-regulation and/or post-filtering.

Best regards,

Charles